The current study examined the relationship of several variables at the patient (n = 2780), physician (n = 166), and group practice (n = 45) levels for predicting receipt of annual mammography screening. Patient-level variables included constructs from the Triandis Model of Choice; physician-level variables included measures of barriers and receptivity to prevention, as well as demographic information. Hierarchical modeling demonstrated that variables at the patient and physician level reliably predict annual mammography screening, while frequency of screening did not vary across group practices after accounting for patient and physician variables. Patient-level predictors included social norms, perceived consequences and perceived barriers. The only physician-level predictor identified was annual mammography recommendation. These findings add to data which emphasize the importance of public education and social support in health maintenance activities.